Inside The Internet

The first message sent across the internet was supposed to be the word LOGIN. The engineer at the keyboard managed to send only two letters before the entire system crashed. The letters were L and O. The first word humanity ever pushed through a global computer network was literally just LO, as in the beginning of hello, or a warning, like lo and behold.Two letters. Then silence. Nobody in that room had any idea they had just nudged open a door that billions of people would eventually walk through every day without even noticing it was there.The engineer who typed those letters, Leonard Kleinrock, was not trying to change the world that night in nineteen sixty nine. He was just trying to get a clunky university computer in Los Angeles to talk to an equally clunky machine at Stanford. The phone line between them hissed and crackled. The computers filled entire rooms and needed their own air conditioning. The users were a handful of researchers with thick glasses and thicker grant proposals. There were no cat videos, no social media, no shopping carts, and no spam. There was only a question that sounded almost childish in its simplicity: can one computer send a message to another computer far away without either of them knowing exactly what route that message will take.That question turns out to be the seed of everything that followed.To understand why Kleinrock’s LO mattered, you have to see what came before it. For most of history, information moved at the speed of horses and ships. A letter from London to New York took weeks. A rumor could die out before it ever reached the next village. The first big upgrade was the telegraph, in the early eighteen hundreds, which squeezed messages into a stream of electric pulses racing along copper wires. For the first time, information outran people. A sentence that once took days to deliver could arrive in minutes.

The telegraph gave way to the telephone, then to radio and television. In each case, the pattern looked familiar. One sender, many receivers. A newspaper publishes. A radio station broadcasts. A TV tower sends out a signal. The information flows in one direction, over a line or through the air, and usually someone owns the channel in between. Telegraph companies owned telegraph wires. Phone companies owned phone lines. Broadcasters owned frequencies. If you wanted to speak to the world, you needed to ask one of those gatekeepers for permission, or enough money to lay your own cable. Information was fast, but it was still controlled.By the nineteen fifties, that control was starting to look dangerous. The Second World War had ended with mushroom clouds. The Cold War hung over everyone like a permanent thunderstorm. Military planners in the United States stared at maps of their long distance telephone network and saw fragility everywhere. The system was built like a tree, with thick trunks of long distance lines splitting into thinner and thinner branches. Knock out a few trunks with bombs and the whole tree died. Messages could no longer cross the continent. Orders could not reach bases. A surprise attack could decapitate the nervous system.In nineteen sixty, a researcher at the RAND Corporation named Paul Baran handed the Pentagon a different kind of sketch. Instead of a tree, his diagrams looked like a spider web, dots linked by short, crisscrossing lines. No central hub. No single vulnerable trunk. Every node was connected to several others, and messages would not follow a fixed path. They would hop from node to node, finding any route that was still intact. Even if half the nodes in the network were destroyed, the rest could still talk.Baran’s idea sounded reckless to traditional communications engineers. Telephone networks were obsessed with reliability and order. When you dialed a number, the network set up a single dedicated path between you and the other person, a continuous wire in effect, reserved only for your call for as long as you talked. It was called circuit switching, and it made sense in a world where conversations were rare and expensive. Baran proposed the opposite. Instead of reserving a path, break every message into small blocks of data, send each block separately, let them mix with blocks from other messages, and reassemble them at the destination. The network would no longer promise a perfect path. It would promise that pieces would eventually find their way.For phone engineers, this was like telling a railway company to melt down its tracks and throw everything into the river, trusting that somehow all the pieces of train, track, and cargo would drift downstream and sort themselves out into their original trains again. It sounded mad. Yet the madness had a hidden logic.The key insight was that conversations, whether between people or machines, do not actually use a line continuously. When you talk on the phone, half the time you are listening. When two computers swap data, there are pauses between bursts. Circuit switching wasted enormous capacity by reserving a whole lane of traffic for each call, even when that lane was empty. Baran’s scheme, called packet switching, turned the whole network into a kind of data highway where every small car, each packet, carried a label saying where it was going. The road did not care who owned the car, only where it needed to end up. Lanes were shared, and the road could route around potholes and fallen bridges.

Packet switching did something else that would turn out to be profound. It separated the intelligence of the network into two layers. The wires and routers knew how to move packets, but they did not particularly care what was inside them. The endpoints, the computers at the edges, knew what the packets meant. That split, between a dumb but flexible transport system and smart devices at the edge, is the skeleton of the modern internet.While Baran was sketching webs for generals, a very different group of people was wrestling with a very similar problem for much more mundane reasons. The Advanced Research Projects Agency, ARPA, inside the American defense department, had started funding computing research at universities around the country. At places like UCLA, Stanford, Utah, and MIT, researchers were building enormously expensive computers to do simulations, graphics, and time sharing. Each machine was a small kingdom. If you were a researcher at one site and wanted to use a program on another, you had to fly, or mail magnetic tapes across the country. The agency that had paid for these toys started to lose patience.The frustration was simple. They had invested millions of dollars in computing power, but each island of silicon was mostly isolated. Larry Roberts, an ARPA program manager, wrote to his sites with a blunt proposal. Instead of asking the agency for more computers, researchers should first connect the ones they already had and share them. It was not a military command and control project. It was a cost saving measure and a scientific convenience. The first network that would become the internet began as an attempt to let academics borrow each other’s machines.In nineteen sixty nine, four universities prepared to be the founding members of ARPAnet, the ancestor of the internet. UCLA. Stanford Research Institute. The University of California at Santa Barbara. The University of Utah. Each site received a refrigerator sized box called an Interface Message Processor, built by a small contractor. These IMPs were the early routers. They spoke packet switching to each other and spoke whatever weird language each local computer understood on the other side.On the evening of October twenty ninth, nineteen sixty nine, Kleinrock’s team at UCLA sat down to connect through their IMP to the one at Stanford. They dialed the phone company line linking the two IMPs. They checked voltages on oscilloscopes. Finally, one programmer typed the L of LOGIN. At Stanford, the machine recorded it. He typed the O. Stanford received it. Then he typed the G, and the remote system promptly crashed. The log from that night is memorable in its understatement. It shows two entries. The first says we sent the L. The second says we sent the O. The third notes that the system crashed.Two characters were all the network could handle on its first try. That failure was its first success. It proved that packets could go from one university to another, across a messy phone line, without anyone needing to reserve a dedicated circuit. Soon, LOGIN would work. Then dozens of sites would join, then hundreds.ARPAnet was tiny at first. In nineteen seventy one, it had only about fifteen nodes. By nineteen seventy five, it was still only around sixty. The people using it all knew each other’s names. They published lists of every host on the network as a kind of shared phone book. There were no anonymous trolls, no influencers, no commercial services. There was no world wide web. Yet within this small community, you can already see the outlines of how the internet would reshape human life.One of the first applications anybody wrote for ARPAnet was not some grand scientific tool. It was email. A programmer named Ray Tomlinson, working on how to send files from one machine to another, realized that messages between users could ride along the same mechanism. He picked the at symbol, already on the keyboard but rarely used, to separate the user from the machine name in an address. User at host. It felt like a minor convenience at the time. In hindsight, it was a new way of describing where a human could be reached in a world where place was no longer a physical office or street, but a slot on a network.As more universities joined the network, more little applications like that bloomed. A programmer wrote a tool called finger that showed who was logged into remote machines and what they were working on. Others created discussion boards and shared directories. They were not trying to invent social media. They were simply scratching their own itches, trying to make collaboration easier when collaborators no longer had to share a building to share a project.Beneath this gentle chaos, however, a deeper technical problem was brewing. ARPAnet had shown that packet switching worked. Other networks, built by other groups, began to appear. The French built Cyclades. The British built the National Physical Laboratory network. Companies like Xerox built internal networks. Each one used its own rules for addressing machines and routing packets. They were like isolated city transit systems, each with its own tickets, timetables, and track gauges. If you wanted to move a packet from one network to another, you had to build an awkward custom bridge.

The question slowly shifted from how to build a packet switched network to how to make many different networks behave like a single, seamless whole. In the mid nineteen seventies, two researchers, Vint Cerf and Bob Kahn, proposed an answer that would become the heart of the internet. They called it Transmission Control Protocol and Internet Protocol, TCP and IP. Their idea was a layer on top of existing networks that would hide their differences and make them all look the same to the computers using them.They gave every machine on every participating network a numerical address. They defined how big packets should be, how to number them, how to acknowledge receipt, and how to request a resend if something was missing. Most importantly, they insisted that the middle of the network should not need to know about the paths between every possible pair of computers. Each router would only know about its neighbors and a few general routes. If a path failed, routers would learn to send packets elsewhere. The edges of the network, the sending and receiving computers, would take responsibility for putting the stream back together in the right order.This did something almost philosophical. It made the network agnostic. A packet did not care if it crossed copper wire, fiber optics, satellite links, or light pulsing through glass under the ocean. It did not care if the information inside was a letter, a video, or the instructions for a bank transaction. The network just moved packets from one numeric address to another, the same way a postal service moves envelopes without reading every letter.On January first, nineteen eighty three, at midnight, ARPAnet officially switched from its older protocol to TCP IP. It was not a glamorous moment. There were no countdowns in public squares, no fireworks. On most of the planet, nobody noticed anything at all. But inside the handful of rooms housing ARPAnet routers, the change meant that their small american research network had just become part of something broader, something defined not by who owned the wires, but by which machines agreed to speak the same language of packets and addresses.Only in hindsight does that night look like a hinge. Before it, networks were local fiefdoms. After it, they could federate. In principle, anyone, anywhere, with any kind of link, could build a network and connect it to the growing internet, as long as they followed the shared rules. The core power moved from hardware to protocol, from cables and switches to a kind of agreed upon grammar for data.Even with that grammatical revolution, though, the internet of the nineteen eighties was still a very small town. It was mostly academics, government researchers, and a growing number of hackers who loved exploring systems just to see what they could make them do. To use it, you needed obscure command line tools. Addresses were numerical. Finding a resource meant knowing exactly which machine it lived on. Every journey online felt like walking through a dark library with a flashlight, guided by handwritten notes from friends.Two innovations would change that feeling dramatically. The first solved a naming problem. Humans are terrible at remembering long strings of numbers, but good at names. You probably do not know your friends’ phone numbers by heart anymore, but you remember their names in your contact list. The early internet had the opposite problem. Machines knew each other only by their numeric addresses, and people passed around host files that mapped names to numbers by hand.To untangle this mess, researchers created the Domain Name System, DNS. DNS works like a global, distributed phone book. Instead of memorizing that a server at a university has the address one hundred twenty eight dot thirty two dot ten dot eight, you just remember berkeley dot edu. When you type that name, your computer quietly consults layers of DNS servers, asking, in effect, who is in charge of dot edu, who knows about berkeley within that space, and what number is associated with that name. The answer comes back in milliseconds. You never see the numeric address. You see a human friendly label.DNS also did something sneaky and powerful. It created a hierarchy of control. At the top were a handful of root servers that knew about the major domains, like dot com, dot org, and country codes like dot uk or dot jp. Below them, organizations could manage their own piece of the naming tree. This arrangement kept naming from becoming a centralized choke point while still avoiding chaos. It was possible for anyone to register a name, but you could not, for instance, grab google dot com if someone already had it.The second innovation turned the internet from a set of message pipes into a place that felt like it had rooms and pages. In nineteen eighty nine, at CERN in Switzerland, a British physicist named Tim Berners Lee was frustrated by how hard it was to share documents between machines. Scientists there used different systems, formats, and computers. Reports sat in scattered file systems and databases. Newcomers had no easy way to discover what existed. Berners Lee wanted something almost childlike in its directness, a web of linked documents you could browse by following your curiosity.

His proposal described three simple ingredients. First, a way to identify documents with addresses that could include not just a machine but a path within it, which became the Uniform Resource Locator, the URL. Second, a language to describe how those documents should look and which parts of them linked elsewhere, which became Hypertext Markup Language, HTML. Third, a protocol to transfer those documents on demand, which became Hypertext Transfer Protocol, HTTP. Together, they formed the world wide web.Hypertext itself was not a new idea. People had been dreaming of linked, non linear documents for decades. What Berners Lee did was marry that dream to the existing internet and, crucially, make his system open. He did not patent it or charge license fees. Anyone could set up a web server and anyone could write a browser that could read pages. When he released the first browser and server in the early nineteen nineties, only a few hundred websites existed. The web was a toy used by physicists. Yet its logic was addictive.A web page could contain text, images, and links. Click on a link and the browser would fetch a different document, perhaps from a different continent, seamlessly. You did not need to know the path your requests took. You did not need to log into each remote computer manually. The browser and the underlying protocols did the hard work. For the first time, a person with no technical background could wander the internet by following their interests, moving from page to page as easily as turning pages in a magazine.Browsers improved quickly. In nineteen ninety three, a group at the National Center for Supercomputing Applications released Mosaic, one of the first widely used graphical browsers. Mosaic could display images inline with text, an apparently minor feature that made the web feel alive rather than academic. Within a year, commercial versions appeared, and the first primitive websites for newspapers, shops, and fan communities launched.Then something subtle but decisive happened. Until the early nineteen nineties, the backbone of the internet in the United States was still funded by government research networks, which mostly banned commercial traffic. In nineteen ninety one, that ban was lifted. Private companies began building their own internet backbones. Internet service providers started selling dial up connections to households. The sound of a modem screeching over a phone line became the audio logo of the era. Entities that had nothing to do with science or the military, from bookstores to bands to local chess clubs, created web pages.In less than a decade, the number of websites exploded from hundreds to thousands to millions. The internet stopped being a tool for a narrow elite and started turning into a mirror, however imperfect, of humanity’s obsessions. Recipes and conspiracy theories, academic papers and fan fiction, pornography and prayer circles, stock tips and support groups, all jostled for space on servers that now lived in office closets, data centers, and eventually anonymous warehouse sized farms.This transformation created a new problem that might be the most human of all. With so much information available, how did anyone find anything. Early web directories, curated by hand, tried to list interesting sites in categories like a digital phone book, but they quickly drowned in the flood. What the web needed was not a librarian, but a map that could redraw itself every minute.Search engines provided that map. Early ones like Archie and Veronica indexed file names and titles. Later ones, like AltaVista, started reading entire pages. The most revolutionary twist came from a student project at Stanford called Google, which took advantage of a fact baked into hypertext itself. Every time a page links to another, it is quietly voting that the other page is worth visiting. By counting and weighing these votes, Google’s algorithm could rank results not just by whether they contained certain words, but by how central they seemed to the web’s own structure.Search engines turned the web from an overwhelming sea into something people could navigate in seconds. You did not need to know the address of a weather site. You just searched for weather and clicked the first few results. Underneath, fleets of software crawlers, called spiders, roamed the web day and night, following every link they could find, building an index of billions of pages. The internet became not just a place you visited, but a place that watched and learned from your visits.By the early two thousands, another layer was forming on top of this already thick stack. At first, the web had mostly been read only. People consumed pages others created. Making your own site required some technical skill. As tools improved, that barrier fell. Blogging platforms let anyone publish essays in minutes. Photo sites let people share images with friends. Forums and message boards flourished.Then came services designed specifically to capture and amplify social connections. Friendster, MySpace, Facebook, Twitter, YouTube, and countless others turned up in rapid succession, each promising an easier way to connect with other humans through the network. The phrase social media had not existed when Kleinrock typed LO. Within thirty years, it would describe a multi trillion dollar ecosystem sitting on top of the same basic packet switching machinery.

The logic of these platforms was simple and seductive. If the internet connected machines, social media connected identities and attention. Every like, share, and follow became a data point. Algorithms learned which posts kept people staring and which made them look away. Within a few years, your experience of the internet stopped being a more or less random walk and started being a heavily guided tour, optimized to keep your eyes from wandering.Underneath the shifting surface, however, the core technical principles that emerged in that first era remained stubbornly the same. Packets still hopped from router to router, taking whatever path was available. DNS still turned names into numbers. TCP still broke messages into pieces and reassembled them. HTTP still ferried page like content. New protocols for streaming video and real time chat layered on top, but they obeyed the same physics. Engineers swapped copper for fiber, dial up for broadband, and eventually wires for radio waves in wifi and cellular networks. The fundamental idea that any device, anywhere, can speak the same language of packets if it wants, did not change.What did change, dramatically, was scale. In nineteen seventy three, there were a few dozen machines on ARPAnet. In nineteen ninety three, there were about two hundred sites on the web. By the early twenty twenties, estimates suggested that more than five billion humans used the internet in some form, and tens of billions of devices were connected. Refrigerators, security cameras, wristwatches, cars, factory robots, and farm sensors poured data into the same vast fabric that once carried only academic emails.This scale shift has consequences that are easy to miss because they unfold quietly. When only universities used the internet, a glitch meant a few students could not log in to a remote machine. When a hospital network goes down today, surgeries can be delayed and medical records may become unreachable. When undersea cables are cut, entire countries can feel like someone has turned down the volume on reality. When a major cloud provider has an outage, thousands of small businesses discover that the computers they thought they owned are actually just terminals talking to distant machines they never see.The internet did something subtle to geography as well. It did not erase it, but it bent it. Before networks, the distance between two people mattered absolutely. To talk, you needed to share a room, a village, or at least a continent wide system of letters and telegraphs. Today, a programmer in Lagos can collaborate in real time with a designer in Lima and a marketer in London without thinking about time zones until a meeting runs late. A rumor can appear on a message board in one country and trigger protests on the other side of the world by evening.This bending of distance is more than a curiosity. It changes who can participate in conversations that shape culture and power. When an activist in a repressive regime uploads a video of a protest, they are not just documenting for local authorities. They are potentially speaking to millions abroad. When a teenager in a small town discovers a niche community that shares their interests, they are no longer trapped in whatever social roles their immediate surroundings offer. The internet stitches together people who would never have met, for comfort and for conflict.That stitching has a darker edge. The same tools that let people find their tribe can also trap them in echo chambers. Recommendation engines learn what you already agree with and feed it back to you. False information can spread as fast as true information, often faster, because it does not have to wait for facts. States and corporations can watch the trails you leave behind with every click, building profiles that know more about your habits than your closest friends do. A network built to route around damage can also route around traditional gatekeepers, for better and for worse.Through all of this, one quiet truth keeps resurfacing. The internet is not a cloud in the mystical sense. It is a very physical thing. It lives in glass fibers as thin as hair, bundled together and laid across ocean floors by ships that spool cable like thread. It lives in racks of humming servers stacked in windowless buildings that consume enough electricity to power small cities. It lives on towers that scatter signals over neighborhoods and satellites that whisper packets down to mountaintops. Every click you make travels through places, owned and maintained by people and institutions with their own interests.Most of the time, that physicality is invisible. You open an app, and words, pictures, and videos appear. Somewhere, routers look up addresses, DNS servers resolve names, and packets leap across national borders. Lawyers debate data sovereignty while gamers argue about lag. The gap between the simplicity of the surface and the complexity beneath is part of the magic. You do not need to know how it works to use it. Yet understanding at least its outlines reveals something about why it behaves the way it does.

The decision to make the network’s middle simple and general, moving oblivious packets instead of tailor made circuits, is why the internet keeps swallowing new kinds of communication without needing permission from a central authority. Email, the web, streaming video, voice calls, virtual reality, and whatever comes next all flow as different flavors of packets. No one had to redesign the whole thing from scratch when video became important. Engineers just wrote new protocols on top of the old ones.The choice to keep the protocol specifications open and publicly documented is why the internet grew faster than its closed rivals. Anyone with enough patience and hardware could build a piece of the network and plug it in. That openness also meant nobody could easily stop the spread of things they did not like, whether that was controversial political speech or new businesses that threatened old ones. The same neutrality that kept the network undecided about packet contents also made it stubbornly hard to censor completely.This does not mean the internet is some perfect utopian engine. Far from it. The very features that make it resilient also make it hard to govern. You cannot simply point to one office and say this person controls the internet. Standards bodies propose, companies implement, governments regulate at their borders, hackers probe for weaknesses, users invent uses nobody anticipated. The result is more like an evolving ecosystem than a finished machine.Look back at that first LO on the screen in nineteen sixty nine and it is tempting to see destiny in it, as if the network was always going to become what it is now. That is hindsight’s favorite trick. The people in that lab were not dreaming of global e commerce or social networks. They were trying to get a research project to work. Every step since then has been a mix of intention and accident. Baran’s cold war resilience maps, ARPAnet’s academic sharing, TCP IP’s unifying grammar, DNS’s naming tree, Berners Lee’s web of documents, the rise of search, the explosion of social platforms, the spread of mobile phones that turned every pocket into a portal.Remove any one of those and the timeline bends sharply. If packet switching had lost to traditional circuit switching, the cost of connecting millions of small devices might have been prohibitive. If TCP IP had been proprietary instead of open, we might have a patchwork of incompatible global networks. If the web had been patented, its growth would have slowed to a crawl. If search engines had failed to tame the flood of pages, the web might have drowned in its own abundance. If mobile data had remained a luxury, vast regions of the world would still be offline.Instead, those particular dominos fell as they did. The result is the nervous system you are using right now to hear this. When you tap your phone, you are taking part in a conversation that began with two letters and a crash. Your messages, photos, and late night questions join an unimaginably huge river of packets that never sleeps, flowing through a mesh so dense no single mind can comprehend it in detail.The internet connects billions of machines, but what it really connects is expectations. It has trained entire generations to assume that answers are seconds away, that news is instantaneous, that distant friends are always within reach, that their words can leap borders as easily as electrons cross switches. When those expectations meet the limits of infrastructure, law, or politics, sparks fly. The next chapters in the internet’s story will be written in those sparks.

Episode Summary

Two Letters

Old Gatekeepers

ARPA Nets

TCP/IP Moment

Web Awakening

Scale & Impact

Inside The Internet

Episode Summary

Two Letters

Old Gatekeepers

ARPA Nets

TCP/IP Moment

Web Awakening

Scale & Impact

Loved this episode?

Chapter Summaries

Two Letters

Old Gatekeepers

ARPA Nets

TCP/IP Moment

Web Awakening

Scale & Impact

Decision Networks

Inside The Internet

Episode Summary

Two Letters

Loved this episode?

Chapter Summaries

Two Letters

Old Gatekeepers

ARPA Nets

TCP/IP Moment

Web Awakening

Scale & Impact

Decision Networks

Old Gatekeepers

ARPA Nets

TCP/IP Moment

Web Awakening

Scale & Impact

Decision Networks